DocumentCode
3575066
Title
Communication Optimal Least Squares Solver
Author
Kumar, Pawan
Author_Institution
Competence Center for High Performance Comput., Fraunhofer Inst. of Ind. Math., Kaiserslautern, Germany
fYear
2014
Firstpage
316
Lastpage
319
Abstract
For matrix with full column rank, QR algorithm is among the best approach to solve wider class of least squares problem (LS). Using the communication optimal variant of TSQR, we study the scalability of the least squares solver with multiple right hand sides. The communication for TSQR based LS solver for multiple right hand sides is still optimal in the sense that no additional messages are necessary compared to TSQR. However, LS has additional communication volume, and flops compared to that for TSQR. The scalability of the proposed method is studied up to few thousand cores using global address space programming framework (GPI) and pthreads.
Keywords
least squares approximations; mathematics computing; matrix algebra; message passing; GPI; LS problem; QR algorithm; communication optimal TSQR; communication optimal least squares solver; communication volume; flops; global address space programming framework; least squares problem; matrix algebra; multiple right hand sides; pthreads; scalability; Binary trees; Libraries; Message systems; Programming; Q-factor; Scalability; Synchronization; Least squares; PGAS;
fLanguage
English
Publisher
ieee
Conference_Titel
High Performance Computing and Communications, 2014 IEEE 6th Intl Symp on Cyberspace Safety and Security, 2014 IEEE 11th Intl Conf on Embedded Software and Syst (HPCC,CSS,ICESS), 2014 IEEE Intl Conf on
Print_ISBN
978-1-4799-6122-1
Type
conf
DOI
10.1109/HPCC.2014.55
Filename
7056759
Link To Document